EOL Ranking Structure Ideas

[gimp]

I'd be surprised if this has never been addressed. But ranking everybody solely by how many battle wins they have obviously doesn't rank any one person better than the next at all. Play a million battles and win a thousand of them and your ranked 15 on the current ranking with a 0.1 nab percentage win rate (extreme example obviously).

I'm surprised 2nd place, 3rd place etc. in a battle has absolutely no relevance at all. I understand that a complex formula would have to be in order. But with enough columns of data and functions etc. i could see a better ranking structure happening.

Ideas:
What if the ranking was updated monthly?

I'd think a point system similar to tournaments would work well, also giving a heavier amount of points depending on how many people were in the battle. For example winning a battle will always give you the same amount of points but second place in a 5 person would give you 800 where second in a 20 person would give you 950 etc.

Of course you wouldnt want to make it all about absolute points because it would rely a lot on simply playing all the time like million battle nab player analogy above. So another variable would be needing to average out the number of points by battles played. You could reward the player with some small amount of points for playing a certain number of battles per month maybe >100 = 100 points, >200 = 200, etc. (points given here could also take into account how many players were played against?, how well you did etc.)
But the averaging would make it so mass amount of playing but still being a nab wont rank you above someone you shouldnt be.

Now a problem could arise with someone who won some 5 out of 5 balles and then quit, as this doesnt have much confidence behind there skill level, but theyd have a perfect score. Perhaps some >15 battles needs to be played or something for the month before we can rank you. i havent thought of a better way to offset this dilemma yet. but inactivity could diminish your points, if you havent played a battle for a year, then you probably shouldnt be in the top10 anymore so it would be significant loss of points.


For example gimp plays 100 battles may 2015.
we can say 1100 is a perfect score. 1000 being %100 and the extra 100 given for 100 battles played, 90 points for 90 battles played etc. perhaps cap the extra credit at 100?

in a 5 person battle the ranking goes like this

1. zero 1000
2. marrku 800
3. zweq 600
4. gimp 400
5. mawane 200
(for instructional purposes only)

or some shit like this.

then 10 person battle difference by 100s.

i play 40 5 person getting 4th in some 40 of them. and 40 10 person battles getting 8th in 40 of them. i place 1st in 5 of each of them. my total points would be 42,000. then divide by battles played (100) and voila 420 and add 100 for 100 extra credit points for battles played = 520.

when next month i get maybe some 600 points we just average the thing out. 520 + 600 / 2 = 560. Again next month maybe i am amazing and get 1100 now 520 + 600 + 1100 / 3 = 740 not too shabby.

bottome line:
somebody like Leek who has become very good at battles will not make it into the top10 anytime soon because he needs to win a shitload more battles to get there. however he is currently maybe a top10 battler. this is the problem with this absolute wins structure. it seems like a lazy way to do it because nobody could think of anything better.

im sure there are many flaws with my idea but you get it, (maybe somebody with a better brain than mine knows what to do). this would be a lot more exciting and add potentially more activity, people would want to see the site monthly etc. you could even add the top20 ranks as ctrl + f5 or something so we could check it every month. some games like counter strike have had very good ranking structures for their servers very similar to this and updated monthly and i just think eol could have the same thing if some thought was put into it.

[Igge]

This has been discussed in numerous topics before. I'm still a fan of something similar to this idea by mila, as it won't lead to insane amounts of points in the end:

http://mopolauta.moposite.com/viewtopic ... 38#p134138

But I agree, fixing ranking should be one of the top priorities for EOL. Easy for me to say though, as I won't do anything to make it happen except maybe donate a little money to compensate the poor bastard who has to do it (sorry kopa :*).

[gimp]

Yes! mila has better brain than me so lets put his idea here as well.

it was like this
lets say players on 1.-4. position had ranks A,B,C,D before
after balles ranks will be
A(n+1) = A(n) * (1 + k*exp(q*(B(n)-A(n)))) * (1 + k*exp(q*(C(n)-A(n)))) * (1 + k*exp(q*(D(n)-A(n))))
B(n+1) = B(n) / (1 + k*exp(q*(B(n)-A(n)))) * (1 + k*exp(q*(C(n)-B(n)))) * (1 + k*exp(q*(D(n)-B(n))))
C(n+1) = C(n) / (1 + k*exp(q*(C(n)-A(n)))) / (1 + k*exp(q*(C(n)-B(n)))) * (1 + k*exp(q*(D(n)-C(n))))
D(n+1) = D(n) / (1 + k*exp(q*(D(n)-A(n)))) / (1 + k*exp(q*(D(n)-B(n)))) / (1 + k*exp(q*(D(n)-C(n))))
there will be also some max/min value of exp(..) so it wont be any unstable
these rules would keep A(n)*B(n)*C(n)*D(n) constant
if someone much better beats someone who sax, it will increase his coefficient by some 1.000001 and decrease the others one byt same 1.000001, if someone much worse beats someone, ez increase weight a bit more, constants q,k will be some 0.01 or smth, simply somehow set (it means the speed of ranks changing), starting value will be some 1000 or 1.000 or smth, goodplayers will have >1, bad players <1
its not decided if 0 apples results are taken, if same times means same position or not and if same apples means same result (definitely for 0 apples)

[SveinR]

Why not just implement a copy of the ELO system (the one used in chess)? I suppose mila's system might be a version of that and should work well also. I think in the past people didn't want that because that meant you would lose points by performing badly, which might make some people hesitate to join battles. But I doubt this will be much of an issue nowadays.

[Lousku]

Ultimate EOL Ranking discussion

2sec lauta search -,-

[Kopaka]

Why not just implement a copy of the ELO system (the one used in chess)? I suppose mila's system might be a version of that and should work well also. I think in the past people didn't want that because that meant you would lose points by performing badly, which might make some people hesitate to join battles. But I doubt this will be much of an issue nowadays.

One issue with that is that chess is 1on1 while elma battles is an unidentified number of opponents. Which means the way the mila system works (which was implemented for a while) you can gain a lot more points from a battle with 30 players than in a battle with 10 no matter who you're playing against. Another issue was the constants we used giving people too many points, so you could jump up and down really quickly, but that could be fixed with proper testing. As mentioned in the thread lousku linked stini worked on a system, I don't know if it accounted for these things, but he never seemed to finish and I got tired of asking. If anyone has an idea to help this issue I can attempt applying it to mila's system.

[SveinR]

One issue with that is that chess is 1on1 while elma battles is an unidentified number of opponents. Which means the way the mila system works (which was implemented for a while) you can gain a lot more points from a battle with 30 players than in a battle with 10 no matter who you're playing against.

Is this really a problem? The way I see it you should gain more points from winning a battle of 30 players than a battle of 10 players. If a battle actually has that many players there's probably a high chance that it has a fair number of skilled opponents anyway.

[gimp]

Why not just implement a copy of the ELO system (the one used in chess)? I suppose mila's system might be a version of that and should work well also. I think in the past people didn't want that because that meant you would lose points by performing badly, which might make some people hesitate to join battles. But I doubt this will be much of an issue nowadays.

If not ELO there has got to be some other kind of ranking system out there that would be similar yet has multiple players per event yes?

[Mats]

Halo 3 had a great system, where top half gets better rank, and bottom half gets lower ranking if theyre previously ranked better than anyone in top half. Also if anyone beats someone with better rank they get more bonus. This system results in people rank according to skill level. Nowadays most people want better ranking regardless of skill, which makes it pointless really. When I played halo I maybe got to skill 40 and stopd playing. Then when I started after a break, I played worse and maybe went down to 35 before improving again.

[analcactus]

my too cvnts is that winning rows should be taken into account and !rec system maybe too

[SveinR]

If not ELO there has got to be some other kind of ranking system out there that would be similar yet has multiple players per event yes?

You could easily construct a multiplayer-system from ELO by pretending a 10-man battle consists of multiple 1-on-1 matches. Then calculate the end score of each player after calculating the result of each 1-on-1 match and adding these. See for instance this calculator and try some example: http://elo.divergentinformatics.com